THE PALOMAR/MSU NEARBY STAR SPECTROSCOPIC SURVEY. III. CHROMOSPHERIC ACTIVITY, M DWARF AGES, AND THE LOCAL STAR FORMATION HISTORY 1 John E.

Transcription

1 The Astronomical Journal, 123: , 22 June # 22. The American Astronomical Society. All rights reserved. Printed in U.S.A. E THE PALOMAR/MSU NEARBY STAR SPECTROSCOPIC SURVEY. III. CHROMOSPHERIC ACTIVITY, M DWARF AGES, AND THE LOCAL STAR FORMATION HISTORY 1 John E. Gizis Department of Physics and Astronomy, University of Delaware, Newark, DE 19716; gizis@udel.edu I. Neill Reid Space Telescope Science Institute, 37 San Martin Drive, Baltimore, MD 21218; and Department of Physics and Astronomy, University of Pennsylvania, 29 South 33d Street, Philadelphia, PA ; inr@stsci.edu and Suzanne L. Hawley Department of Astronomy, University of Washington, Box 35158, Seattle, WA ; slh@astro.washington.edu Received 21 November 16; accepted 22 March 1 ABSTRACT We present high-resolution echelle spectroscopy of 676 nearby M dwarfs. Our measurements include radial velocities, equivalent widths of important chromospheric emission lines, and rotational velocities for rapidly rotating stars. We identify several distinct groups by their H properties and investigate variations in chromospheric activity among early (M M2.5) and mid (M3 M6) dwarfs. Using a volume-limited sample together with a relationship between age and chromospheric activity, we show that the rate of star formation in the immediate solar neighborhood has been relatively constant over the last 4 Gyr. In particular, our results are inconsistent with recent large bursts of star formation. We use the correlation between H activity and age as a function of color to set constraints on the properties of L and T dwarf secondary components in binary systems. We also identify a number of interesting stars, including rapid rotators, radial velocity variables, and spectroscopic binaries. Key words: Galaxy: formation solar neighborhood stars: chromospheres stars: formation stars: low-mass, brown dwarfs On-line material: machine-readable tables 1. INTRODUCTION M dwarfs are the dominant stellar component of the Galaxy by both number and mass. With main-sequence lifetimes much longer than the age of the universe, they are a fair tracer of the overall properties of the Galactic disk. Their chromospheric activity decays on timescales of billions of years, providing an age indicator that is relevant for studies of Galactic evolution. By relating the activity levels in M dwarfs to age, we can measure the local star formation history. The latter parameter is one of the major requirements in modeling the local substellar mass function (Reid et al. 1999). Moreover, M dwarf components in multiple systems provide constraints on the age of the system. Finally, an extensive sample of young, low-luminosity stars in the field can furnish a prime hunting ground for imaging surveys designed to find young, luminous brown dwarf and giant planet companions. As with other low-luminosity objects, detailed observations are only possible for M dwarfs in the immediate solar neighborhood. The most extensive source for such objects remains the preliminary version of the third Nearby Star Catalogue (Gliese & Jahreiss 1991, hereafter pcns3). Papers I (Reid, Hawley, & Gizis 1995) and II (Hawley, Gizis, & Reid 1996) in this series describe our moderate-resolution (3 Å), red (62 75 Å) spectroscopic observations of candidate M dwarfs in the pcns3. We have used those spectra, together with data from the literature, to estimate distances and spectral types. In Paper I, we defined a volume-limited sample of northern ( > 3 ) stars and investigated the luminosity function and kinematics of lowmass stars in the Galactic disk. Paper II presented data for the southern stars and investigated various aspects of the chromospheric behavior of the whole sample using the H line as a diagnostic. 2 In this paper, we present echelle observations of M dwarfs from the volume-limited sample defined in Paper I. Our echelle spectra are of high (.2 Å) resolution and cover the wavelength range Å for all stars, with data extending to 37 Å for a subset of the brighter stars. These observations encompass chromospheric emission lines due to hydrogen, helium, and ionized calcium. Previous surveys of activity in M dwarfs have generally concentrated on earlier spectral types (M4), as in the H surveys of 22 stars by Stauffer & Hartmann (1986), or have been limited to relatively small samples, such as the 24 late-type M dwarfs observed by Giampapa & Liebert (1986). The only published analysis based on a volume-limited sample is the recent work by Delfosse et al. (1998, 1999a), who present observations of field M dwarfs within 9 pc. Their spectroscopy has a higher resolution than our data, but their sample includes only 118 stars. Our survey therefore provides a more detailed study than heretofore possible of the distribu- 1 Observations were made at the 6 inch telescope at Palomar Mountain, which is jointly owned by the California Institute of Technology and the Carnegie Institution of Washington In both these papers, the tables were printed incorrectly. Combined tables are available in electronic form at the ADC and CDS as catalog III/198.

2 NEARBY STAR SPECTROSCOPIC SURVEY. III tion of chromospheric activity among late-type dwarfs and, combined with an age-activity relation calibrated by M dwarfs in open clusters, the first opportunity to use those stars to probe the Galactic disk star formation history. The following section presents the basic analysis of our spectroscopic data, including radial velocity and rotational velocity measurements and line-strength determinations. Since we have multiple observations of many stars, we have used our data to search for velocity variations, and x 3 identifies several previously unrecognized binary systems. Chromospheric activity is discussed in x 4, where we investigate the H properties of the sample, describe an ageactivity relation calibrated with open cluster observations, and use that relation to probe the recent star formation history of the Galactic disk. Section 5 summarizes our main conclusions. 2. OBSERVATIONS 2.1. The Sample Our primary sample consists of the volume-complete sample of 499 single M dwarfs and M dwarf primaries defined in Paper I. We refer to this as the VC sample. The stars have absolute magnitudes in the range 8 M V 16, declinations north of 3, and distances within completeness limits ranging from 22 pc at M V =8.5.5 to 5 pc at M V = The latter values were derived based on photometric and trigonometric data available in late 1995, coupled with the (M V, TiO5) relation derived in Paper I. Since then, Hipparcos-based trigonometric parallaxes (Perryman et al. 1997), accurate to 1 mas, have become available for approximately two-thirds of those systems. As will be discussed in Paper IV of this series (Reid et al. 22), the addition of the new astrometric data affects the inclusion/ omission of only 15% of the stars in the Paper I sample. While the new distance determinations should be taken into account in the analysis of the kinematics of the local stars (Paper IV), they are of little importance for the analysis of chromospheric activity and age that is undertaken in this paper. Thus, we retain the VC sample as our reference here Echelle Spectroscopy Our data were obtained with the echelle spectrograph (McCarthy 1985) on the Palomar 6 inch (1.5 m) telescope, which has a 2 pixel resolution of 19,. Observations between 1994 May and 1995 February used the original quartz 6 cross-dispersing prisms, giving wavelength coverage from 37 to 95 Å. Beyond 7 Å, there are gaps in wavelength coverage where the orders extend off the CCD. Because our targets are all red stars, very few counts were obtained in the blue ( <48 Å), except for the very brightest stars. The result was that the useful part of the spectrum was squeezed into the lower quarter of the CCD, leading to partial overlap between the reddest adjacent orders. To solve this problem, beginning in 1995 June, a new set of cross-dispersing prisms (with SF3 glass and a 42 apex) was used. This setup gives wavelength coverage from 48 to 95 Å and has complete order separation, although the gaps in wavelength coverage remain. For both setups, the exposure time was 6 s per star, except for the faintest (V > 13) stars for which the exposure times were increased to as much as 18 s. Data were extracted using the FIGARO echelle software (A. Tomaney & J. McCarthy 1996, private communication). Wavelength calibration was determined from a 3 s exposure of a ThAr lamp taken at the beginning of each night. After each observation of a target star, a short (45 s) exposure of the arc lamp was taken at the same telescope position in order to remove instrumental flexure. The spectra were not flux-calibrated Radial Velocities Radial velocities were determined by cross-correlation against reference M dwarfs from Marcy & Benitz (1989, hereafter MB89). The latter velocities are accurate to better than.23 km s 1, a factor of 5 higher than the accuracy of our own observations (as discussed further below). For the bright, early M dwarfs (TiO5 >.5), we used the standard echelle FIGARO cross-correlation program. Each order was correlated with the velocity standards, and the average radial velocity from all the orders was determined. The arc lamp exposures, taken adjacent to each program star observation, were also cross-correlated, providing a correction for flexure. This procedure did not provide reliable velocities for faint, late-type M dwarfs. The lower signal-to-noise ratio at blue wavelengths led to a higher potential for bias from the effects of telluric absorption and night-sky lines. We were able to obtain reasonable results for those stars by individually computing the cross-correlation for each order and combining those measurements to find the median radial velocity. This procedure was used for all stars with TiO5.5. Figure 1 plots the rms dispersion about the mean radial velocity for all M dwarfs with at least four measurements. The distribution suggests a typical internal accuracy of d1.5 km s 1. We can determine our external errors by comparison with previous high-accuracy velocity studies of M dwarfs. The results are shown in Table 1, where ref is the formal uncertainty of the reference sample. We note that a comparison between Tokovinin (1988, 1992) and MB89 (13 stars) gives an rms of only.46 km s 1 but a mean difference of.78 km s 1, consistent with the offset derived from our observations, which are tied to the MB89 system. Similarly, Stauffer & Hartmann (1986) adopt a velocity of 14.1 km s 1 for the velocity standard Gl 526, while MB89 measure a velocity of 15.7 km s 1 for this star. Again, the offset is consistent with our measurement. Finally, we list two comparisons with the Source TABLE 1 Radial Velocity Comparison N ref (km s 1 ) hv rad i (km s 1 ) (km s 1 ) Bopp & Meredith Delfosse et al a Stauffer & Hartmann Tokovinin 1988, Upgren & Caruso Upgren & Harlow a Excluding the three stars discussed in the text.

3 3358 GIZIS, REID, & HAWLEY Vol TABLE 2 Radial Velocities NN Name Julian Date v rad hv rad i V 7... Gl 2 2,45, Gl 4A 2,449, Gl 4B 2,449, GJ 12 2,45, ,45, ,45, ,45, G242-48A 2,449, GJ 15B 2,45, ,45, ,45, ,45, ,45, ,45, Gl 12 2,449, Gl 14 2,449, ,45, Gl 15A 2,449, ,45, Gl 15B 2,449, GJ 11A 2,449, GJ 25 2,45, ,45, Gl 21 2,45, Note. Table 2 is presented in its entirety in the electronic edition of the Astronomical Journal. A portion is shown here for guidance regarding its form and content. Fig. 1. Observed of the measured radial velocities for stars with multiple observations. Early M dwarfs (TiO5 >.5) are shown as open squares, and mid M dwarfs (TiO5.5) are shown as solid triangles. There is no trend with the apparent magnitude of the star. recent observations by Delfosse et al. (1998) since the dominant contribution to the residuals comes from three stars: Gl 26 (V Del = 8. km s 1 ; D = V P6 V Del = 9.2 km s 1 ), G165-8 (8.; D = 15.5), and Gl 268.3B ( 6.; D = 8.4). All three stars are known binaries, while G165-8 is also a very rapid rotator (see below). In general, the comparison indicates that the velocities derived from our echelle spectra are accurate to <1.5 km s 1. We can also examine the quality of the radial velocities found from our previous moderate-resolution spectra. Based on an external comparison with the MB89 standards, we estimated an accuracy of 15 km s 1 (Paper I). After excluding double-lined spectroscopic binaries, velocity standards, and three stars with anomalously large residuals, we have 582 stars in common. The standard deviation is 17 km s 1, confirming our previous estimate. The individual velocities and Heliocentric Julian data for each observation, together with the mean value and the rms dispersion based on our observations, are given in Table 2. In addition to the star name, we list the NN number cited in Papers I and II. While the latter is merely the rank order in pcns3 and is not an officially recognized designation, it provides a straightforward means of cross-referencing results between tables for this relatively large data set. We also include measurements of three M dwarfs that are not included in either the pcns3 or Papers I and II but that are discussed in Gizis & Reid (1997). Both the primary G and the secondary LP were on the slit, so the reported measurement is for a blend of the two stars. G is a short-period, double-lined dme system Line Strengths To investigate chromospheric activity, we measured the equivalent widths of atomic features using an automated program that counted the flux in rectangular passbands. The adopted line (from F1 to F2) and pseudocontinuum (PC1 to PC2 and PC3 to PC4) regions are given in angstroms in Table 3, and the average values observed for each star are given in Table 4. The appearance of the emission lines in a strong dme star are illustrated in Figure 2. For those sufficiently bright stars observed with the old echelle prisms, the bluer Balmer emission lines were also measured. The NaD lines were not measured because of contamination by sky emission and the difficulty in defining a pseudo-continuum. The measurements of the He 6678 Å line were difficult because of the weakness of the feature and the nearby TiO absorption band. We found that none of the stars had detectable lithium absorption, as expected for M dwarf stars with ages exceeding 5 Myr. Figure 3 compares the H equivalent width measured from the echelle observations with the lower resolution measurements from Paper I. For the early M dwarfs, the 3 Tables 2, 4, and 5, together with the data tables from Paper I and Paper II are available from our Palomar/MSU Web site at

4 No. 6, 22 NEARBY STAR SPECTROSCOPIC SURVEY. III TABLE 3 Emission Lines Name PC1 PC2 F1 F2 PC3 PC4 He D He H H H H H H higher resolution observations are systematically smaller by.5 Å, probably reflecting differences in the placement of the pseudo-continuum. The large scatter for the cooler dwarfs is due to the variability of these stars s chromospheres. For most purposes, it is more useful to compare line fluxes rather than equivalent widths. Absolute calibration is not available for all of the stars in our sample, but the continuum flux near H can be derived from broadband R C photometry (Reid, Hawley, & Mateo 1995). To improve the calibration and extend it to H, we obtained low-resolution spectrophotometry of a subset of 15 stars using the McCarthy spectrograph with a 15 line grating. The observations were made under photometric conditions at the Palomar 6 inch on 1996 July 9 and 1. The wavelength coverage was Å with a resolution of 6 Å, providing a good match to the echelle observations. In Figure 4, we show that use of the Bessell (199b) filter response curves with our spectra reproduces the sixth-order polynomial relation between R I C and V I C derived by Bessell (199a). The deviations seen at the blue end correspond to stars earlier than type M and therefore do not effect our results. Using these low-resolution spectra, we find that the continuum flux at H and H can be derived from broadband photometry using the following linear relations: 2:5logF H ¼ð21:68 :4Þþð:974 :38ÞR C ; 2:5logF H ¼ð21:1 :6Þþð1:55 :55ÞV C ; where the fluxes are measured in ergs s 1 cm 2 Å 1. These relations are applicable for stars with types M M6.5. They are used to transform the measured emission-line equivalent widths to the flux values discussed in x 4.1. We note that differences in continuum placement at low and high resolution might lead to small systematic errors. V-band photometry is available for all the M dwarfs in our sample. In cases where R C photometry was unavailable, we estimated the V R C color using the TiO5 value V R C ¼ 3:2345 8:9529 TiO5 þ 12:41 TiO5 2 5:6274 TiO5 3 : When V I C colors were needed but unavailable, we used V I C ¼ 6: :676 TiO5 þ 17:6957 TiO5 2 8:934 TiO5 3 : Again, these relations are applicable for stars with spectral types M M6.5. TABLE 4 Equivalent Widths NN H He D3 He 6678 H H H H H6 N H H H Note. Table 4 is presented in its entirety in the electronic edition of the Astronomical Journal. A portion is shown here for guidance regarding its form and content.

5 336 GIZIS, REID, & HAWLEY Vol HeD3 8 He NaD Fig. 2. Spectra of Gl 285 (NN 1219) showing the H,H, He D3, NaD, and He 6678 Å emission lines 2.5. Rapid Rotators 5-5 K7 M M1 M2 M3 M4 M Fig. 3. Comparison between the H emission line strength measured at high resolution in this paper and in our previous, low-resolution work (Paper II). Points where the current measurements exceed 5 Å are shown as circles; weaker dme stars are shown as triangles. A systematic difference of.5 Å is evident, which we attribute to differences in placing the pseudocontinuum. The increased scatter among later type stars is probably due to variability (see x 4.1 and Fig. 7). TiO5 Rotational broadening is measurable for a handful of stars in our spectra. For each observing run, we artificially broadened a star of known low rotation (v sin i < 2kms 1 ) using a rotational broadening function (Gray 1992). We adopted a limb-darkening coefficient of.6, shown to be an appropriate value by Marcy & Chen (1992). However, measuring v sin i by cross-correlation against the standards proved ineffective, probably because of small mismatches in spectral type. Instead, we emulated the method of Marcy & Chen (1992) by using 2 fitting to determine the best fit to individual atomic lines. Our lower resolution requires that we use relatively strong atomic lines, as in the region Å. Detections are limited to v sin i >2kms 1 (3 pixels). We estimate a measurement accuracy of 5kms 1. Stars with detected rotational broadening are listed in Table 5. Our results are quite repeatable; for example, G (NN 3453) was observed six times, each time yielding v sin i between 35 and 4 km s 1. The most notable star is G165-8 (NN 2128), which has 13 measurements that are all consistent with v sin i =8kms 1, similar to the extreme rotator Gl 89 (Pettersen et al. 1987). One of our observations clearly shows H emission from a secondary companion, while others show asymmetries in the line profile. The rotational velocity is close to that of the fastest rotating M dwarfs in Pleiades (Terndrup et al. 2), suggesting that the system is relatively young, d 1 8 yr.

6 No. 6, 22 NEARBY STAR SPECTROSCOPIC SURVEY. III Fig. 4. Synthetic colors for M dwarfs using our low-resolution spectra and the filter curves of Bessell (199b). The solid curve is the Bessell (199a) polynomial fit to actual M dwarf photometry. We conclude that our spectra are correctly flux-calibrated. A caveat to all of our rotation measures is that close binary pairs that happened to have velocity differences of 2 km s 1 at the time of our observations could be misidentified as rapid rotators. For example, we measure v sin i =3kms 1 for Gl 735 (NN 2976), but we exclude the star from Table 5 since it is identified as SB in the pcns3. Gl 26 (NN 933) and Gl 829 (NN 336), both known double-lined spectroscopic binaries (SB2 s), also show significant line broadening in our spectra. Confusion is less likely where multiple measurements are available. NN TABLE 5 Rapid Rotators Name v sin i (km s 1 ) G GJ 154A G LHS Steph G G Gl BINARIES Table 6 lists 23 dwarfs that are SB2 s. These systems must have relatively short periods, but our data provide insufficient temporal coverage to attempt period determinations. Note that most objects in the present sample were observed only once, so additional double-lined systems that were in an unfavorable configuration at the time of our observation remain undetected. Thirteen of these 23 SB2 systems have published orbits, while the other 1 are new discoveries. A few single-lined stars with multiple velocity measurements, both our own and those available in the literature, have velocities spanning a larger range than expected given the individual uncertainties. This may reflect the presence of an unseen companion (i.e., SB1 system). Table 7 lists stars where the overall rms dispersion of the observations exceeds 4kms 1. Other stars listed in Table 2 with >3kms 1 may also merit additional observations. Table 8 lists other candidate SB1 systems, where a significant velocity difference exists between our observation and the literature data. With potential velocity amplitudes of a few kilometers per second and periods of a few years, these systems are likely to have separations of 1 AU, equivalent to >1 for stars in this sample. These are therefore good candidates for highresolution imaging (speckle, adaptive optics, or interferometry) and radial velocity monitoring at higher resolution. Several stars have known companions that may be responsible for the observed velocity differences. Poveda et al. (1994) have published a catalog of candidate wide binary or multiple systems in the pcns3 catalog. We have radial velocity measurements for several of their candidates, including two with separations exceeding.1 pc and therefore of potential interest for Galactic dynamics. At such separations, orbital velocities are less than 1 km s 1,so TABLE 6 Double-lined Binaries NN Name Reference Gl GJ 154A LTT Gl Gl Gl 278C GJ 269A G Steph Gl Gl G LTT LHS Gl 63.1A GJ 123AC Gl Gl 815A Gl 866A Gl 867A GJ G G References. (1) Strassmeier et al. 1993; (2) Tokovinin 1992; (3) Delfosse et al. 1999a.

7 3362 GIZIS, REID, & HAWLEY Vol. 123 TABLE 7 Candidate Single-lined Binaries: Multiple Observations NN Name Reference LP Gl Gl 58A G Gl 867B 2.9 References. (1) Delfosse et al. 1999a; (2) Harlow we expect the radial velocities of the components to agree within our measuring uncertainties. Table 9 shows the comparison, where the final column gives the observed velocity difference in terms of our measuring uncertainty, which we take as 1.5 km s 1. Of the two wide systems, Gl 469 and Gl 471 are clearly not associated (unless one has an unseen companion), but our data show only a 3 difference between Gl 48 and the Gl 22 system. Except for Gl 14 A/C, where orbital motion might affect the velocity of the brighter star, the remaining candidate binaries have velocity differences consistent with our measuring uncertainties. 4. CHROMOSPHERIC ACTIVITY, AGE, AND THE STAR FORMATION HISTORY The chromospheric age-activity correlation among mainsequence stars has been studied extensively for solar-type dwarfs, where it is usually parameterized in terms of the t to the half law (Skumanich 1972), FðCaÞ ¼At 1=2 ; where F(Ca) is the Ca ii K emission-line flux, A is a constant, and t is the age. Based on that calibration and observations of local G dwarfs, several groups have attempted to reconstruct the recent star formation history in the disk. In particular, both Barry (1988) and Rocha-Pinto et al. (2) claim that the data indicate several significant bursts of star formation over the last 4 Gyr. On the other hand, Soderblom, Duncan, & Johnson (1991) argue that these bursts are the result of a more complicated activity-age relation than that implied by a simple Skumanich-type power law. Our goal is to use our observations of M dwarfs in the VC sample to address the local star formation history. To that end, we first investigate the range of activity and the correlation of activity with age for the low-mass M dwarfs. In particular, we do not require that the M dwarfs follow the same age-activity relation used for the G dwarfs Activity The primary indicator of chromospheric activity in M dwarfs is H emission. In Figure 5, we plot the equivalent width of H as a function of TiO5 (spectral type/effective temperature). Figure 6 is an expanded view of the absorption (negative) equivalent width portion of Figure 5. Active stars known to be short-period binaries (Tables 6 and 7) are shown as open circles, while members of the VC sample are shown as solid triangles. Additional stars from this paper that are not in the VC sample are shown as open triangles. To aid discussion, we have labeled five groups (A E). Group A. The majority of M dwarfs show H absorption, with larger average absorption equivalent widths at early spectral types. H absorption does not necessarily indicate an absence of chromospheric activity, as discussed by Cram & Mullan (1979) and Cram & Giampapa (1987). These authors showed that the presence of a weak to moderate chromosphere induces H absorption in M dwarfs, which enhances the photospheric absorption. Only a rather strong chromosphere will produce H emission. Thus, M dwarfs with the strongest H absorption probably have moderate chromospheres, while those with no chromosphere will exhibit only weak (or perhaps no) absorption K7 M M1 M2 M3 M4 M5 M6 TABLE 8 Candidate Single-lined Binaries: Velocity Differences NN Name Dv (km s 1 ) Source Gl 228A Gl 319A Gl Gl Gl Gl References. (1) Upgren & Caruso 1988; (2) Upgren & Harlow 1996; (3) Stauffer & Hartmann TiO5 Fig. 5. H equivalent width as a function of TiO5 (spectral type). Stars in the VC sample are shown as solid triangles, stars with emission due to a close companion are shown as open circles, and other non-vc stars are shown as open triangles.

8 No. 6, 22 NEARBY STAR SPECTROSCOPIC SURVEY. III TABLE 9 Candidate Wide Binaries Star 1 V rad (km s 1 ) Star 2 V rad (km s 1 ) Separation (AU) DV/ Gl G22C Gl 22B 3.4 Gl Gl Gl 14A Gl 14C G G Gl Gl Wo 949A Wo 949C Gl 72A Gl 72B Gl 897A....8 Gl see discussion of group E. The observed width of the absorption sequence in Figure 6 is larger than our measuring uncertainty and therefore probably reflects real star-tostar variations at a given effective temperature. These variations are presumably correlated with the level of chromospheric activity. In addition, the fraction of stars per spectral type bin that are in group A decreases toward later spectral types, which is probably related to the increase in the lifetime of strong chromospheric activity for lower mass stars (see discussion in x 4.2). Groups B and C. Most of the early (M M2.5) dwarfs with emission, including all the early M active binaries (BY Dra class) in the VC sample, cluster near the upper envelope of observed activity. We have marked these dwarfs as group B. There is a striking lack of early M dwarfs with weaker emission, indicated as group C. This behavior was noted by Herbst & Miller (1989), who point out a similarity with the Vaughn-Preston gap in the Ca ii emission of G dwarfs K7 M M1 M2 M3 M4 M5 M TiO5 Fig. 6. H absorption as a function of TiO5 (spectral type). Symbols are as in Fig. 5. (Vaughan & Preston 198; Henry et al. 1996). Such a gap can be interpreted either as a stellar population (age) or a stellar chromosphere effect. In the former case, one assumes that age and H emission have a simple relationship, and the distribution reflects a lack of intermediate-age (group C) stars compared with younger (group B) and older (group A) stars. In the latter interpretation, some aspect of chromospheric physics makes it unlikely that a star is observed with intermediate-strength emission lines. For example, stars may remain in the group B state for a long time, then rapidly evolve through the group C region. Group D. The observed distribution of H equivalent width changes at TiO5.5 (spectral type M2.5/3). Among earlier spectral types, emission is limited to 2.5 Å above the mean level of absorption; for the mid M (M3 M5.5) dwarfs, the emission-line equivalent widths are distributed roughly uniformly between Å and a maximum value of 6 1 Å. This change in properties was also noted by Herbst & Miller (1989) on the basis of data for fewer stars. Little structure is evident, although there is a suggestion of a clump at H 4.5 Å, TiO5.4. According to the empirical mass spectral type relations derived by Kirkpatrick & McCarthy (1994), spectral type M3 corresponds to.26 M, while M2 corresponds to.37 M. This range encompasses the mass at which M dwarfs are expected to become fully convective (.3 M ; Burrows et al. 1997). The change in the distribution of activity strengths may be related to changes in the magnetic dynamo generation mechanism and/or the atmospheric structure among fully convective stars (Hawley, Reid, & Gizis 2). Group E. A few dm stars have unusually weak H absorption. They are sometimes called dm(e) stars, since they appear to be intermediate between the dme and dm stars. Two different states contribute to this group. Some stars may have extremely weak, or possibly no, chromospheres. The absorption line, or lack thereof, then simply reflects the weak photospheric absorption that disappears at later types. Others have moderately strong chromospheres that have begun to produce enough emission to fill in the absorption line, but not enough to produce a fullfledged emission line. Examples of both types are known (Byrne 1993; Doyle et al. 1994). Two of the stars in group E, GJ 162 (NN 647) and LHS 64 (NN 338) are old, metalpoor M subdwarfs (Gizis 1997), and as such are likely to have weak, if any, chromospheres. Further study of group E stars may be useful to characterize this rare, inactive state among M dwarfs.

9 3364 GIZIS, REID, & HAWLEY Vol. 123 An interesting question raised by our analysis is whether there is an evolutionary pattern through the groups we have identified. Thus, for example, an early-type M dwarf might start out quite active (group B). As the chromosphere begins to weaken with age, the H emission may fade rapidly according to the particulars of the line radiative transfer leading to a quick passage through groups C and E. Further chromospheric weakening may lead to a slower evolution from the top (less absorption) of group A to the bottom of group A, and finally, as the stars age even further (approaching the age of the Galactic disk?), the chromosphere begins to disappear altogether, with the evolutionary path progressing back through group A to an endpoint in group E. For later type stars, it may be that the radiative transfer in the H line is such that it is easier to drive the line into emission. Thus those stars start and remain in group D throughout most of their evolution, at least to their current ages. We are pursuing these ideas by searching for other chromospheric indicators that would allow us to distinguish between moderate activity (filled-in H) and no activity (weak or no H absorption) and by developing chromospheric models that predict the behavior of the H (and other) emission lines depending on parameters that characterize the strength of the chromosphere (Hawley et al. 22). We obtained monitoring observations of the nearest M dwarfs as part of this program. These allow us to examine variability in the emission line strength. Figure 7 plots the observed standard deviation in equivalent width as a function of emission line strength for stars with at least four observations. For dme dwarfs with H d 5 Å(including all early M dwarfs and the weaker mid M dwarfs), variations are d.5 Å, with typical variations at the 15% level. In contrast, stars with stronger emission are significantly more variable, with variability exceeding 3%, corresponding to changes of several angstroms in equivalent width. As discussed extensively elsewhere (e.g., Reid et al. 1995), equivalent width measurements are useful in segregating stars within a relatively limited range of effective temperature, as in defining the groups in Figure 5, but do not provide a good measure of the absolute level of activity because of variations in the continuum flux as the effective temperature changes. We have therefore used the relations defined in x 2.4 to transform our equivalent width measurements to line fluxes and formed the ratio of the line luminosity to the bolometric luminosity an absolute measure of the activity strength for each star. Figure 8 plots the result for H as a function of TiO5 (equivalently color, spectral type, effective temperature, and mass). The group B stars are clearly evident at spectral types earlier than M2, while the group D stars comprise the cooler half (smaller TiO5) of the diagram. It is interesting that binaries (open circles) are found scattered throughout the latter group, with no clear clustering at the most active levels. The discussion of the groups given above is applicable also to this figure, with the group B stars showing high and rather uniform activity and some evidence for binaries having stronger activity, while the group D stars have a much larger scatter. The mean level of log (L H )/L bol ) is roughly 4., in agreement with the value we found for the Hyades (Reid et al. 1995) and in our lower resolution survey of these same field stars (Paper II). Figure 8 makes it clear, however, that the scatter among the group D stars (later types) is not uniform. There are more stars that show activity above the mean level (between 3.5 and Fig. 7. Observed standard deviation ( H ) in the H equivalent width as a function of EW H for dme stars with at least four measurements. Stars with EW H >5Åexhibit significantly larger scatter. 4), but the stars below the mean level have a larger range (between 4 and 5). With reference to our previous speculative discussion about chromospheric evolution, this could be evidence that the younger, most active stars cluster around log (L H /L bol ) 3.8.2, while the older, less active stars spread out according to their age and initial activity level. It is notable that the difference in scatter between group B and group D is still very striking, which may mean that some change in the dynamo generation at the fully convective boundary is also required (along with evolution and the particulars of H line transfer) for a full physical explanation of these phenomena. Figure 9 shows the Balmer decrement, the ratio of H to H flux. A typical value in the Hyades is 4, somewhat larger than the mean found here of 2.5 (which does agree with our result in Paper II). There is a slight downward trend toward later types among the group B stars and again a notable increase in the scatter, particularly toward larger values of the decrement, among the group D stars. It may be significant that this scatter is apparently restricted to the later spectral types among group D. The scatter is reminiscent of that seen in Figure 7 in H. There are eight stars shown with H > 1.5 km s 1, and three of these (NN 172, 1724, and 1326) have the three highest values of the Balmer decrement, while four of the remaining five (NN 24, 23, 1398, and 1934) have decrements well above the average, between four and 1. A very strong Balmer decrement and significant variability probably indicate that we have

10 No. 6, 22 NEARBY STAR SPECTROSCOPIC SURVEY. III K7 M M1 M2 M3 M4 M TiO5 Fig. 8. Chromospheric activity level expressed as the ratio of the H luminosity to the bolometric luminosity. Symbols are as in Fig Fig. 1. He D3 emission as a function of H emission. Symbols are as in Fig. 5. observed flaring activity in these stars; the decrement actually decreases in flares on earlier M dwarfs (e.g., AD Leo, Hawley & Pettersen 1991) but is known to increase strongly during flares on later M dwarfs (e.g., VB 1, Herbig Fig. 9. Ratio of the H to H emission. Symbols are as in Fig ; 2MASS J149, Liebert et al. 1999). The final star, LHS 1723 (NN 855), is anomalous in both figures, lying at about (1, 1.7) in Figure 7 and at about (.4, 1.2) in Figure 9. This star has very weak H emission compared with H and also shows far more variability than other weak H emitters. It will pose an interesting challenge for chromospheric models. Finally, our observations comprise the most extensive available set of helium D3 (5876 Å) emission-line data for active M dwarfs. Figure 1 shows that there is a strong correlation between the He D3 and H emission equivalent widths, with the slope of a simple linear fit of.12. Alternatively, the flux in the He D3 line may be predicted from the H flux by F HeD3 =.48F H. The good correlation suggests that both features probably originate from the same region within the chromosphere. This is in contrast with previous observations and modeling of Ca ii, Mgii, and H emission in M dwarfs, which implied that the emission in those lines was coming from different chromospheric regions (Mauas & Falchi 1994; Giampapa, Worden, & Linsky 1982). An extensive study of He D3 in G and K stars by Saar et al. (1997) suggested formation in the upper chromospheres of those stars and good correlation of the He D3 flux with those of Ca ii and C iv (similar to our result here for H). Andretta & Giampapa (1995) have outlined a method using He D3 observations to infer the filling factor of active regions in F and G stars. It would be of great interest if such a technique was applicable also for M dwarfs.

11 3366 GIZIS, REID, & HAWLEY Vol An Age-Activity Calibration For M Dwarfs Calibrating age-dependent relations obviously demands stars with well-determined ages. In the Galactic disk, open clusters give an independent age determination, and observations of chromospheric activity levels of cluster members provide the calibration. The traditional approach, as in Barry (1988), is to determine a relationship between the activity level (e.g., equivalent width or other measure) and the age. By applying this law to each field star, unique ages can be derived. While we have proposed above a possible qualitative evolutionary sequence that might eventually allow a unique age determination for each M dwarf in our sample, we are far from certain that our proposal is correct and have not even begun to determine a calibration. In addition, observations of clusters show that there is some spread in activity at a given age. Moreover, as demonstrated in Figure 7, variability is clearly present in individual M dwarfs, complicating any effort to determine an exact activity level. Finally, the appearance of Figure 5 suggests potential inconsistencies in applying this method: If the lack of group C stars is interpreted as a lull in star formation, where is the corresponding gap in group D? If the concentration of stars in group B is a burst of star formation, why is there no corresponding concentration at the top of group D? Hawley et al. s (1999b) analysis of M dwarfs in open clusters suggests a different approach. While the activity levels of stars in a given cluster exhibit considerable scatter, there is a well-defined V I C color at which activity becomes ubiquitous. All stars redder than this color are dme (defined as EW H 1. Å), while the bluer stars are dm without emission. This effect was commented on originally by Stauffer et al. (1994) for Pleiades. Observations currently exist for M dwarfs in six clusters: IC 262 and IC 2391 (3 Myr; Barnes et al. 1999), NGC 2516 and Pleiades (125 Myr; Stauffer, Schultz, & Kirkpatrick 1998), the Hyades (625 Myr; Perryman et al. 1998), and M67 (4. Gyr; Dinescu et al. 1995). Hawley et al. (1999b) have used those observations to determine the relationship between the H limit color and the age in years, V I C ¼ 6:91 þ 1:5ðlog ageþ ; as shown in Figure 11. We can use this correlation to transform an observed distribution of chromospheric activity as a function of V I C color to an estimate of the cumulative age distribution. We define f dme as the fraction of dme dwarfs at a particular color, (V I C ) i, f dme ¼ NðdMeÞ=NðdM þ dmeþ ; ðv I C Þ¼ðV I C Þ i : ð1þ This ratio provides an estimate of the fraction of stars younger than age i, where, from equation (1), log i ¼½ðV I C Þ i þ 6:91Š=1:5 : Thus, for example, the fraction of dme dwarfs at (V I C ) = 1. is directly proportional to the relative number of stars that have formed in the last yr, while f dme at (V I C ) = 2.6 corresponds to the relative number of stars with ages of less than 1 Gyr. In this manner, we can determine f dme for a range of V I C color and map the cumulative star formation history of the Galactic disk V-I = (log Age) IC 2391, IC 262 NGC 2516, Pleiades Hyades log Age (yrs) Fig. 11. Calibration of V I C at the H limit as a function of cluster age from Hawley et al. (1999b). See text for discussion Star Formation History of the Galactic Disk We have applied the analysis technique outlined in the previous section to data for the M dwarfs in the VC sample. For stars lacking V I C measurements, we use the TiO5 index to estimate the color. The results are plotted in Figure 12, giving f dme as a function of the inferred age. A constant star formation rate (i.e., f dme directly proportional to age) is shown for reference as the thick solid line with slope unity in this figure. Two different ways of computing the fraction of dme stars are illustrated. The long-dashed line (connecting solid triangles) gives the fraction found by weighting all stars equally. The thin solid line (connecting open triangles) shows the fraction found by weighting each star by the inverse of its W velocity, as proposed by Wielen (1974, 1977). The latter approach allows for the increased scale height of higher velocity (older) stars and consequent shorter residence time in the solar neighborhood. However, this method adds statistical noise by placing significant (undue?) weight on a small number of high-velocity stars. In both cases, we exclude the SB1 and SB2 (short-period binary) systems, since the activity in those stars may be influenced by other effects than age. Figure 12 indicates that the overall star formation history is broadly consistent with a constant star formation rate. The major feature notable in the distribution is a step at 1 Gyr that, if taken at face value, would indicate that 1% of the local disk stars formed in a burst at that time. However, this feature is unlikely to be real, since it corresponds with the group B/D transition in Figure 5, which we discussed at length in x 4.1. Such a distinct change in activity properties is unlikely to be well described by a simple linear relation. Data for clusters with ages between the Hyades (.6 Gyr) and M67 (4 Gyr) would be useful in confirming whether this feature has an astrophysical, rather than evolutionary, origin. The other characteristic of the star formation history illustrated in Figure 12 is a slight deficiency in the number of young (less than 1 Gyr) stars relative to the number of older (greater than 1 Gyr) stars. At ages of less than 1 Gyr, both the weighted and unweighted solutions match the expectation of constant star formation (slope unity), with slopes of and.95.5, respectively. For the full age distribution, however, we derive best-fit slopes of M67

12 No. 6, 22 NEARBY STAR SPECTROSCOPIC SURVEY. III Log Age (yrs) Fig. 12. Star formation history of the solar neighborhood (assuming a Galactic disk age of 1 Gyr). Our M dwarf results are shown for the W-weighted (solid line, open triangles) and unweighted (long-dashed line, filled triangles) analyses. The thick solid line, slope unity, illustrates the expected distribution for a constant star formation rate. The closed circles on this line mark the ages of the calibrating open clusters. Barry s (1988) G dwarf results (short-dashed line, crosses) are also shown together with the results from the Rocha-Pinto et al. (2) G dwarf analysis (dotted line, open circles). The M dwarf results do not show the excess of young stars suggested by the former analysis and also fail to match the details of the latter for the unweighted solution and for the weighted solution. This result implies a slight decrease in the star formation rate in recent times. However, this could reflect incompleteness in the VC sample, which, since it is derived primarily from proper-motion surveys, may be deficient in young, low space-motion dwarfs. Figure 12 also compares our analyses with the star formation histories proposed from G dwarf studies by Barry (1988) and Rocha-Pinto et al. (2). The poor time resolution at large ages, coupled with the lack of calibrating clusters and the relatively small size of the VC sample, limits the utility of comparisons at older ages (greater than 3 Gyr). Nonetheless, it is encouraging that all of the models predict similar fractions of stars younger than 4 Gyr. At younger ages, the M dwarf data are better calibrated. There is no evidence from our analysis for the substantial numbers of young G dwarfs ascribed by Barry (1988) to a recent burst of star formation. Similarly, our data fail to match the details of the Rocha-Pinto et al. (2) analysis. Indeed, Rocha-Pinto et al. s burst A of young (d.5 Gyr) G dwarfs is accompanied by a deficiency of M dwarfs, while their AB gap at 1 2 Gyr corresponds to an apparent excess of M dwarfs (although, as we noted above, we believe this feature reflects a deficiency in our analysis method rather than a burst of star formation). The AB gap found by Rocha-Pinto et al. (2) corresponds to the Vaughan-Preston gap (Vaughan & Preston 198). As noted above (x 4.1), Herbst & Miller (1989) have suggested that the sparse number of early-type M dwarfs with weak emission (our group C) might be an analogous feature. However, observations of Hyades M dwarfs show that the H limit corresponds to TiO5.55 in that cluster. Thus, the weak emission M dwarfs in group C must have ages of less than the age of the Hyades (.6 Gyr) i.e., ages that match Rocha-Pinto et al. s (2) burst A. As Soderblom et al. (1991) and Rocha-Pinto et al. (2) have discussed, one needs either a complicated G dwarf ageactivity relation to match a constant star formation history or a complicated star formation history to save the simple G dwarf age-activity relation. Given the lack of agreement between our M dwarf analyses and the G dwarf analyses, we believe that the complicated G dwarf age-activity relation explanation is favored. The existence of a large spread in activity in the coeval G dwarfs of M67 (Giampapa et al. 2) and the large spread in rotation rates in stars of young clusters (Barnes 1997) suggests that any age-activity relation is complex, with stochastic star-to-star variations. Indeed, M dwarfs show similar behavior, with a smattering of dme dwarfs bluer than the H limit in some clusters. Binarity may well be a contributing factor at both spectral types. Other measures of the recent star formation history have been made without reference to chromospheric activity. Hernandez, Valls-Gabaud, & Gilmore (2) have used Hipparcos color-magnitude diagrams to derive the star formation history of the solar neighborhood within the last 3 Gyr. They find an oscillatory component of star formation with a period of.5 Gyr superposed on an underlying constant star formation rate. Their Figure 4 indicates that they see roughly 2.5 times as much star formation from Gyr as at.3 1. Gyr. It is suggestive that this is similar to the deviations from constant star formation seen in our Figure 12 that is, the possible deficiency of young stars. Neither our data nor Hernandez et al. (2) show the lull in star formation between 1 2 Gyr seen by Rocha-Pinto et al. (2). Our present analysis is only a first step toward using M dwarfs as probes of Galactic star formation. Additional observations of M dwarfs in clusters, particularly clusters older than the Hyades, are required to improve the ageactivity calibration during the important 1 4 Gyr time period. Further observations in clusters and the field are needed to quantify the significance and range of star-to-star variations, and (with sufficient resolution) to provide additional information on the evolution of activity at a given spectral type, both among the group B and group D stars. Finally, a nearby star sample that is both unbiased kinematically and complete over a larger volume will provide improved statistics for the local field stars. We are currently developing such a sample in a project undertaken under the auspices of the NASA/NSF NStars program (Reid & Cruz 22) Brown Dwarfs The ability to constrain the ages of individual M dwarfs can be used to set limits upon the ages of brown dwarf candidates that happen to be in binary systems